Clever Geek Handbook

Thirty reports of DevOops 2019: Tim Lister, Hadi Hariri, Roman Shaposhnik and other stars of the international DevOps

On October 29-30, St. Petersburg will host DevOops 2019, a conference dedicated to DevOps engineering solutions. The main topics are clouds in general, and Cloud Native in particular, observability, monitoring and auditing, CI / CD, security, and so on - in general, everything you can expect from a conference devoted specifically to devops.







This hubpost is a review of the DevOops program, which we wrote together with the conference program committee.







In short:









The program is large, with a total of 30 reports.









Keynotes are led by Tim Lister (co-author of Peopleware), Hadi Hariri (head of Developer Advocacy at JetBrains) and Roman Shaposhnik (member of the board of directors of the Apache Software Foundation and Linux Foundation LF Edge ).







Under the cut, we’ll talk about what’s going on in the world of devops, we’ll break up the events into groups and see how it all fits into a program of 30 reports.







Case study



The first group of reports is Case Study. Recently, the situation in the world has been developing rapidly, for example, GitLab’s war continues with everyone: GitLab has everything, and they are going to sell it as a single product (as stated in a scandalous article by The Register CEO Sid Sijbrandij).







However, not everything is so rosy, and Nikita Sobolev just wants to tell how they moved from GitLab to GitHub and why. In short, GitHub also had everything, albeit in preview, but the same Actions work, Package Registry works, Security also works, and so on. On the other hand, Actions still continue to pump out the entire repository, regardless of which pull request you are testing in which branch. Yes, in the end you will climb onto Jenkins, because life is hard and full of horrors, but at least you can already assemble the docker image, and most of it will be enough for a start. This was one example from the Case Study category, but there were several reports in general:









NB: Baruch Sadogursky continues to make presentations on each new DevOops. Will he hit the top again? We make bets.







Service mesh



In the Service Mesh reports, we explore ways to solve a problem of increasing complexity. We are great, we came up with the idea of ​​splitting monoliths into microservices, but instead of finally solving the problem, we were faced with the incredible complexity of the microservice world. Meshes were invented to make complexity less, but in the end ... what happened turned out. Something tells me that it hasn’t become easier with meshes, but this is a topic for a separate big conversation.







Now you can find more and more articles and reports on the topic: let's stop micro-hardcode and return to the usual monoliths. In fact, as soon as it became clear that microservices reduce complexity in architecture, but increase the load on admins, people came and said: “Oh, maybe this is not always necessary”? This thesis has not disappeared, it is not always necessary. For example, if your system needs to be integrated with something big and external, with some kind of thick database, then the number of problems solved by microservices is less than the problems created - the whole state will go through this database, and microservices are no longer micro, for they cannot live without each other. In the program, we even have such monolith lovers - for example, Alex Thissen with the report “Flagging your features” talks about the fact that you can take one monolithic application, cover it with feature flags and then always roll from the wizard.







But who will fight with service meshes? Ask our speakers!









CI / CD



The old Jenkins can be run without Jenkins! You can run in Travis, anywhere, how do you like this, Elon Musk? (This is quite serious now). In general, since Kubernetes is everywhere now, all our CI / CD tools adapt to this fact, Kubernetes needs to be supported. That's why we got JenkinsX, so new chips appear in TeamCity, that's why GitHubs and GitLabs deploy their CIs - everyone needs Kubernetes.







The advent of Kubernetes has changed the approach to CD. As making the CD easier, new variations of how fun to roll canary deployments, blue-green deployments, and so on - a bunch of ready-made abstractions that you can use and enjoy your life began to appear. An example of a CI / CD built on Cloud Native principles is Tekton . We still have nothing about Tekton (except that it is mentioned in the reports of Oleg Nenashev and Burr Sutter ), but in the spring we will try to do it. JenkinsX is exactly the same chip, created on the basis of Cloud Native projects for Cloud Native. If anyone is interested in what this same Cloud Native is, it is worth reading about 12-factor applications , this is it. As Kelsey recently joked:









CI / CD category reports:









Monitoring



In the monitoring world, everything is much less stormy, but there are fundamental questions. For example, it is often said that no one has learned to monitor monoliths. It seems that the problem is not that - after all there is nothing much to learn. The problem is that most monoliths are legacy, and screwing monitoring to existing applications is a pain. If they tell you now: write a monolith in such a way that it is convenient to monitor it, it’s time to spit: you take everything that we love, starting from logs and metrics and ending with tracing, you write everything beautifully there and get full observability.







The problem is that today we are talking about screwing monitoring to existing large monoliths, and this is rather nontrivial. And when everything works out, you will have to somehow live with the resulting Frankenstein. Therefore, we have a report by Dmitry Stolyarov about the culture of on-call , it is not very technical, but remember, devops is not only about tools! Philipp Krenn will tell you that when we scale, we begin to lose events, and, in general, this is normal, but here the auditors come to us and say - we want to watch individual events! How to marry scaling and auditing is unclear, an unpleasant problem.







In general, we have not yet learned to fasten monitoring to monoliths, and microservices have piled on top of work. Microservices and Cloud Native made us take a completely different look at observability, because we understand that old methods like dumb logging, on which we continue to do cat, stop working. Recently, a joke slipped somewhere on Twitter: “If you paste the ID into five different tools, and then look for them using this ID, then you yourself are an observability tool.” In microservice architectures moving towards the reactive model, observability is built on the events with which they are transferred internally. And if it's an orchestration, then there are no events, and you have to disassemble the logs in a different way. The world has become several times more difficult to observe, and not everyone has learned to observe it.







Reports from the Monitoring section:









Cloud



Cloud is the largest and most voluminous topic. It used to be that public clouds were “our everything.” Then it turned out that not everything. Then it turned out that it wasn’t ours either! Many lovers of private clouds appeared, hybrid clouds arose. It did not start this year, but much earlier. Now one of the main questions is how to combine all this. For example, how VMWare is combined with AWS, because VMWare came to Azure , and also to AWS , and this is already pretty big news all over the year.







Of course, everywhere, in most reports (not just the Cloud section) Kubernetes is mentioned one way or another. He infiltrated everywhere, and someone even starts to wait - when will the Kubernetes killer appear? So far, this is not visible. A couple of years ago, the main question was - how to live with this complex incomprehensible thing, but now everyone is already used to the evil neighbor and have learned to negotiate. Operators? Knative? Kotlin DSL?







This bold new world is so large and diverse that it makes no sense to list everything here, just take a look at this list of reports:









Keynotes



We also have three keynotes. They occupy the largest room available, there are no other reports in parallel with them, and they are designed for a wide audience, they are led by their most famous speakers.







The conference opens with Timothy Lister, co-author of Peopleware, Waltzing with the Bears, Adrenaline junkies and template zombies. All of these books are classics in their field and are written with colleagues from the Atlantic Systems Guild . In the report “Characters, community, and culture: Important factors for prosperity” Tim will talk about best practices for organizations, working culture, useful and harmful aspects of working in a company. In general, about what he has been talking about for decades, but updated to the realities of 2019. If the details are interesting, just recently we did a great interview with him for Habr . Do they write a new book - yes, they do, read the interview.







The first day ends with Hadi Hariri, the legendary leader of the Developer Advocacy team at JetBrains, an open source developer and speaker for over 15 years. In his Removing the barriers report , he suggests thinking about the question: what if all the usual barriers and problems have disappeared, what's next? Does this really lead to increased productivity and guaranteed problem solving? It turns out that not everything is so simple, and the lack of barriers is itself a worthy topic for discussion.







And finally, the conference ends with Roman Shaposhnik, a member of the board of directors of the Apache Software Foundation and Linux Foundation LF Edge , who personally had a hand in the Linux kernel, Hadoop, ffmpeg and other popular projects. His keynote, “Why the IT industry is going through dark times, how DevOps is to blame, and why Capital can help,” will try to answer some philosophical questions about the rise of multimedia cloud platforms, open source platforms (Kubernetes and Cloud Foundry), Edge Computing and so on.







What's next?



The full conference program is published on the site , there are detailed descriptions everywhere, comments of the program committee are everywhere, and tags like #kubernetes allow you to navigate the contents without going to the report card.







We remind you that DevOops 2019 will be held on October 29-30 in St. Petersburg, tickets can be purchased on the official website of the conference . You can learn about all the significant news either from our blog on Habré or by subscribing to the mailing list on the main page .







More articles:


All Articles